Think about the last time you needed to buy something you knew nothing about. A few years ago, you would have typed it into Google and worked your way through a page of blue links. Now, a lot of people just ask. They open an AI chat, type "what's the best X for under £50?", and take the answer at face value.

That shift is happening quickly, and it changes the question every brand needs to ask. It used to be "how do we rank on page one?" Now it is also "how do we become the thing the AI actually recommends?"

If that feels like one more thing to worry about, take a breath. You don't need a new marketing team to get there. You do need to think a little differently about where recommendations come from.

How people are really shopping now

The behaviour change is already well underway. In B2B, it is even more pronounced: recent research found that nearly nine in ten business buyers now turn to generative AI as a first port of call for product and supplier information.

The pattern is simple. A shopper describes what they need in plain language, the assistant compares the options, and hands back a short list of suggestions. Often, the shopper trusts that short list before they visit a single website. So if your product isn't on it, you may never get the click at all. The decision has already been made.

It is a data question, not just a copywriting one

Here is the part that surprises most people. AI assistants don't see your beautiful homepage the way a human visitor does. They pull from structured information: product feeds, specifications, pricing, stock availability, schema markup, and the reviews scattered about the web. ChatGPT, for example, draws a large share of its product answers straight from Google Shopping data.

So the brands that get recommended aren't the ones with the cleverest taglines. They are the ones whose product information is complete, accurate, and consistent everywhere an AI might go looking.

If your size guide is missing, your materials field is blank, or your price on one marketplace contradicts the price on another, you have given the assistant every reason to recommend a competitor it understands better. It isn't being unfair. It just can't vouch for what it can't read.

And here is the thing about messy product data: it is almost never a content problem. It is a systems problem. The gaps and contradictions appear because the information lives in too many places at once.

What AI assistants look for

You don't need to second-guess the algorithms. A few practical things make your catalogue easy for an AI to suggest with confidence:

• Complete product attributes. Dimensions, materials, use cases, compatibility. These are the exact details people put into their questions, so fill the gaps.
• Accurate, real-time stock status. Being recommended for something you can't actually ship is worse than not being recommended at all. It ends in a refund, a frustrated customer, and a review that follows you around.
• Consistent information across channels. Your Shopify listing, your Amazon page, and your own site should all tell the same story.
• Genuine reviews. AI tools lean heavily on real customer sentiment to decide whether to trust a product. Honest reviews do more for you here than any amount of marketing polish.

Why this is really a systems problem

This is where it gets practical for operators. You cannot feed an AI clean, current data if your own systems don't agree with each other.

Picture the typical setup. Stock sits in a spreadsheet. Pricing lives in Shopify. Amazon and eBay get updated by hand. Your accounting tool has its own version of the cost figures. None of these are wrong on their own, but together they never quite line up. There is no single, trustworthy version of the truth to share. Not with a customer, and certainly not with a machine that refreshes its answers constantly and compares you against everyone else in your category.

That is why a unified system matters so much here. When all your product, stock, and pricing data live in one connected place, every channel an AI reads is pulling from the same source. Update a price once, and it is right everywhere. Sell the last unit, and every listing knows within seconds. The assistant gets one clear, consistent signal instead of four contradictory ones, and consistency is exactly what earns its trust.

Agentic commerce raises the stakes again. That is the next step, where AI agents don't just recommend products but go ahead and buy them on a shopper's behalf. Those agents check live availability and pricing through data feeds and APIs before they commit. Stale data doesn't just cost you a recommendation anymore; it costs you the sale, often without you ever seeing the order that didn't happen.

Why a Retail-First ERP system makes a difference

Not every system handles this well. A traditional ERP was built for a slower world and often needs custom code to talk to each new sales channel. A loose stack of apps bolted together leaves gaps for data to fall through. Neither gives you the single, real-time view that AI tools depend on.

A Retail-First ERP like Brightpearl by Sage is built for exactly this. It is omnichannel by design, so it pushes consistent product and stock information out to your retail channels natively and in real time. That is the same clean, current data an AI assistant needs to read you correctly. It works this way:

• One accurate record: product details, pricing, and stock sit in one place and stay consistent across every channel an AI might read.
• Real-time availability: stock levels update as orders come in, so you are never recommended for an item you can't ship.
• Clean data out to your channels: connected feeds to Shopify, Amazon, and your marketplaces keep listings current without anyone re-keying them by hand.

The nice part is that none of this is AI-specific work. You are just keeping your operational house in order, and AI tools happen to reward brands that do.

Where to start

You don't have to crack AI recommendations overnight. Start with the unglamorous work: get your product data complete, get it accurate, and make sure it says the same thing everywhere. 

In practice, that means working from one connected system rather than several that disagree. It is the foundation everything else sits on, and it is the same foundation that keeps your operations calm as you grow into new channels.

Ready to give AI assistants something clean to recommend?